The goal of this proposal is to predict and design specific protein interactions by using a combination of computational and experimental methods. To reduce the complexity of the problem, we target interactions mediated by the coiled coil. The coiled coil is a simple and important interaction motif estimated to occur in roughly 3-5% of all proteins, including many important for human disease. It has been studied extensively over the past 15 years. Consequently, existing knowledge provides a framework in which to attempt the challenging problems of interaction prediction and design. The specific aims of the proposal are: (1) To measure the pairings that occur among coiled coils found in human and yeast bZIP transcription factors. This will be accomplished using a large-scale protein microarray assay. The resulting data will be used to develop computational methods for predicting coiled-coil interactions from sequence. The bZIP interaction screen will also provide a wealth of data for the study of transcriptional regulatory networks that involve human oncogenes, including Fos and Jun. (2) To improve computational methods that have been developed for protein design so that these are more suitable for the problem of designing interaction specificity. 3) To use computationally-guided methods from Aim 2 to design peptides that bind specifically to targeted human bZIP coiled-coil domains, and to test these designs experimentally. This will provide a comparison of designed peptides with naturally occurring ones (Aim 1) that share similar interaction properties. It will also constitute a rigorous test of our basic understanding of coiled-coil recognition, and it will provide useful reagents for perturbing transcriptional regulatory networks. (4) To design a peptide that acts as an inhibitor of the oligomerization of BcrAbl, a human oncoprotein. The coiled coil-mediated dimerization of Bcr is implicated in more than 95% of chronic myelogenous leukemias. Together, these studies will improve our understanding of the molecular basis of protein interaction specificity and provide tools that can be used to rationally alter protein structure and function. The methods proposed can be applied to a wide range of different domain-domain interactions, so the insights that we achieve will have broad significance for the study of protein-protein associations generally.